Martin Knoche comments

Results 6 comments of


                                            Martin Knoche

cosine distance between different people is < 0.5

Indeed, the FaceTransformer + OctupletLoss model, tends to project features more closely than for example ArcFace + Octuplet. You have to set the threshold based on your preferences. I can...

cosine distance between different people is < 0.5

Well - make sure the images have float pixel values ranging from 0 to 1. And make sure your images are stored in RGB format - cv2 for example uses...

Acurracy is pretty low

Hi, it's not the best face detection implemented here. It can happen that it does not detect a face. The models should only download ones. The code is checking the...

Colab demo notebook is not working.

Hi, thank you very much for opening the issue. I didn't update the Colab Book according to the updates in demo.py - that's why there was an error. I fixed...

Over resource limits on Streamlit Cloud

Thanks, I am aware of the issue and will work on a fix soon.

About training

Hi, unfortunately, I am not able to provide you any code for the training anymore. I used PyTorch and Tensorflow 2 for training and simply used the provided code as...